Airport charges and marginal costs for Spanish airports before the process of partial privatization

Transcription

1 Airport charges and marginal costs for Spanish airports before the process of partial privatization PRELIMINARY AND INCOMPLETE Please, do not quote Ramón Núñez-Sánchez 1 Soraya Hidalgo-Gallego 2 Valeriano Martínez-San Román 3 This article aims at comparing the most relevant aeronautical airport charges with their corresponding marginal costs for the Spanish airports in the previous period of the partial privatization process. To that end, a flexible short-run variable cost function system is estimated using a panel of thirty-five airports over a 5-year period. The results show the existence of nonneutral technological change and some degree of overcapitalization for Spanish airports. Moreover, aeronautical airport charges are set above the short-run marginal costs with the exception of the smallest airports. Keywords: airport regulation, airport charges, cost function, Spanish airports. JEL codes: R48, L93, H54. 1 Corresponding author. Departamento de Economía. Universidad de Cantabria. Avda. Los Castros, s/n, Santander (Spain). Phone: nunezr@unican.es. 2 Departamento de Economía. Universidad de Cantabria. 3 Departamento de Economía. Universidad de Cantabria.

2 1. Introduction The evolution of the airport sector in the last 3 years has meant that transport policies have to face new challenges derived from technological and institutional change in the airline sector or the growing importance of airport infrastructures as generators of economic activity. In this sense, airport pricing is considered as one of the most relevant issues for policymakers. However, this phenomenon has not been studied until relatively recent times. Lu and Pagliari (24) point out that one of the most critical issues facing policymakers is whether it is preferable to adopt a single-till, where airport charges are set to cover both aeronautical and commercial activities, or dual-till approach, where only aeronautical activities are taken into consideration. Their research concludes that the single-till approach is desirable where excess of capacity exists. Other authors obtain similar results in later theoretical models such as Czerny (26). The empirical literature that aims at analyzing airport pricing is even more recent. Bel and Fageda (211a) examine different factors determining airport charges for a sample of European airports. Their results show that non-regulated private airports set higher charges compared to those from regulated private or public-owned airports. They also find no differences in the level of airport charges among different regulation mechanisms. Bilotkach et al. (212) use a sample of European airports in order to analyze the determinants of airport charges as well. Their results, however, show that aeronautical charges are lower at airports when single-till regulation is employed. Moreover, privatized airports present lower airport charges. Bottasso and Conti (212) analyses the cost structure of British airports for the period They find that average costs start to increase when airport traffic reaches 14 million passengers. Moreover, two relevant results with regard to the regulation of the UK airport industry arise. First, the evidence of over-capitalization for the largest regulated airports might support the view that the price-cap regime presents common consequences with a rate of return regulation. Second, long-run marginal costs are higher than average costs for the largest regulated airports. This finding might indicate that regulated airport charges based on average cost scheme would be inefficiently low. Mathisen et al. (214) review the charge scheme used to finance airport infrastructure in Norway. They derive long-run marginal costs for passengers and air traffic movements at Norwegian airports using panel data analysis for the period The estimated marginal

3 costs are then used as the basis for a revised airport charge scheme designed to meet the principle of maximizing social benefits. The results suggest that there should be a shift towards a relatively higher charge for passengers compared to air traffic movements. This article analyses marginal costs and aeronautical airport charges for the Spanish airports in a pre-privatization period (29-214). We propose a flexible short-run cost function system to estimate marginal costs for passengers and aircraft operations. The results show that aeronautical airport charges are set above the short-run marginal costs with the exception of the smallest airports. This paper contributes to the airport pricing literature in several ways. From a methodological point of view, the use of maximum take-off weight (MTOW) instead of air transport movement (ATM) as the output variable, or the consideration of outputs variables being endogenous are some novelties regarding previous studies (Bottasso and Conti, 217). Moreover, there is no previous study to our knowledge in which airport charges are compared with their corresponding marginal costs for the Spanish case. Martín et al. (211) estimate different cost function specifications for the Spanish airports between 1991 and They show that economies of scale are not exhausted at any output level, as well as significant technological progress. They also find that the possibility of input substitution is very limited and that input demands are elastic. The rest of the paper is organized as follows. We describe the Spanish airport system and its pricing policy for the pre-privatization period, from 29 until 214, in Section 2. Section 3 presents an econometric specification suitable for studying long-run marginal costs at Spanish airports. Section 4 describes the data set and Section 5 presents the results and derives marginal costs for passengers and MTOW. Finally, we conclude and offer some policy implications derived from our analysis. 2. The Spanish airport system 2.1. An overview of AENA The public firm Spanish Airports and Air Navigation, AENA, was created in 1991 to manage the main Spanish airports on a centralized basis. AENA has been attached to the Spanish Ministry of Development, which, in accordance with the mandate established by the Government, set its action guidelines, approves the annual plan of objectives, and monitors its activity. Other features of the Spanish airports are as follows: (i) the existence of cross subsidies among airports (Bel and Fageda, 211b); (ii) discriminatory airport pricing strategy based only on airport passenger traffic; (iii) lack of individual airport marketing policies related to charges,

4 quality or innovation. In this sense, regional governments use promotion agreements with air carriers in order to stimulate new or existing routes, especially in regional airports. This rigidity provokes lack of transparency in these kinds of agreements (Núñez-Sánchez, 215). (iv) Airports institutional framework, which differs notably from other industries with similar technological characteristics such as ports (Núñez-Sánchez et al., 213), is considered an obstacle to take advantage of potential economic efficiencies (CNMC, 214). In 21, the executive order 13/21 created a new corporation, AENA Aeropuertos, aimed at managing Spanish airports and providing greater flexibility to the system. This new regulation allowed the entry of private capital with a maximum amount of 49%. In addition, it contemplates the creation of subsidiary companies or concession contracts for the individualized management of some airports. Finally, those tools were never developed. The capital of the new firm established in June 211 was completely public whereas the centralized management of airports remained. A new reform in 212 (executive order 2/212) contemplated the gradual transition from a single-till system into a dual-till system in five years through a cost recovery model. AENA's charges should adjust to their market value progressively. This fact would allow an increase of aeronautical revenues for Spanish airports. In July 214, the government approved a new regulatory framework for setting airport charges called document for airport regulation (DORA) to prepare airport regulation before the privatization process. AENA was partially privatized en bloc in 215. The Spanish government maintained 51 percent of total shares and, therefore, controlled the company but listed 28 percent of the operator on the stock market and sold another 21 percent to three private-sector anchor investors: Spanish infrastructure group Ferrovial, British investment fund TCI, and Spain's Corporacion Financiera Alba fund Pricing in the Spanish airport system The structure of Spanish airport charges from 29 to 214 is established under the Air Safety Law 21/23, which considers airport charges as public patrimonial benefits. Therefore, they are not dependent on decisions related to any market principle. As abovementioned, the centralized airport pricing system, based on a single-till rate of return system, set identical 4 See Hidalgo-Gallego et al. (217) for a more detailed analysis.

5 charges for different subsets of airports defined according their passenger traffic. Table 1 shows the classification of the Spanish airports according to the number of passengers into five groups as proposed by AENA. Table 1. Spanish airport classification in 214 Group GROUP 1 GROUP 2 GROUP 3 GROUP 4 GROUP 6 Airports Adolfo Suárez Madrid-Barajas (MAD); Barcelona-El Prat (BCN) Alicante (ALC); Gran Canaria (LPA); Tenerife Sur (TFS); Málaga-Costa del Sol (AGP); Palma de Mallorca (PMI) Bilbao (BIO); Fuerteventura (FUE); Girona (GRO); Ibiza (IBZ); Lanzarote (ACE); Menorca (MAH); Santiago (SCQ); Sevilla (SVQ); Tenerife Norte (TFN); Valencia (VLC) Almería (LEI); Asturias (OVD); A Coruña (LCG); FGL Granada-Jaén (GRX); Jerez (XRY); La Palma (SPC); Murcia (MJV); Reus (REU); Santander (SDR); Vigo (VGO); Zaragoza (ZAZ) Albacete (ABC); Badajoz (BJZ); Burgos (RGS);Córdoba (ODB); El Hierro (VDE); Huesca-Pirineos (HSK); La Gomera (GMZ); León (LEN); Logroño (RJL); Melilla (MLN); Salamanca (SLM); San Sebastián (EAS); Pamplona (PNA); Vitoria (VIT); Valladolid (VLL) Notes: In general terms, airport revenues are classified into three different types: those related to aeronautical infrastructure services; those which finance support services for aeronautical activities (also called non-aeronautical services); and finally, commercial revenues (Salazar, 213). Table 2 exhibits the revenues structure of Spanish airports in 211 before the implementation of any privatization schedule. Regarding the first type of revenue, aeronautical services are provided to airlines to enable aircraft landing and take-off operations, boarding and disembarking passengers and loading and unloading goods. Aeronautical revenues should finance the main airport infrastructures. It represented 65 per cent of the total revenue in 211. Revenues related to passenger services (27.5 per cent) and aircraft landing (22.3 per cent) being the most important. Commercial revenues include concession activities such as supply of car parking, car rental services, food and beverages, duty-free shopping, etc. They represented 26.8 per cent of total revenues. Commercial concessions, car parking and car rental services were the most relevant activities in terms of total revenues. Finally, support services for aeronautical operations (also called nonaeronautical services) include the handling of the aircraft, the passage and the merchandise during and after the stopover, provision of ramp handling services, food services on board, check-in desks, authorization for restricted areas, or the use of waiting rooms. They represented

6 less than 8 per cent of the total revenues in 211. Table 2. Revenues structure of AENA Aeropuertos in 211 Type of revenue Million Subtotal (%) Total (%) Aeronautical revenues Landing Aircraft parking Passenger services Passenger security Cargo Other Subtotal 1, Non-aeronautical revenues Subtotal Commercial revenues Subtotal TOTAL 2, Source: AENA annual report (212) Due to the importance of passenger and landing charges in terms of aeronautical revenues, we will analyze the evolution of these two airport charges from 29 to 214. In order to take into account the charge regulatory setting, we consider the abovementioned airport classification. We also consider an additional category for insular airports due to the particular regulatory scheme. Figure 1 shows that, in the first two years, passenger charges are relatively homogeneous, with the exception of insular airports. However, differences between groups arise for the next years, especially for the hub airports (group 1). Passenger charges increased more than 2 percent in Madrid-Barajas and Barcelona-El Prat from 21 to 212, consistent with the first proposed privatization schedule, keeping more stable during the last years of the sample. Regarding other groups, their respective passenger charges have kept particularly steady throughout the sample period, except for groups 4 and 5, the smallest airports, for which passenger charges fall notably in 211. This last fact may be explained, at least in part, by a strategy of attracting low cost carriers to operate in these small and regional airports. Figure 2 presents the evolution of the landing charges for each group of airports. As in the case for passengers, landing charges increased for the bigger airports (groups 1 and 2), while they decreased for the rest of the groups. Insular airports show the most stable pattern.

7 16 Figure 1: Evolution of passenger charge by airport group Group 1 Group 2 Group 3 Group 4 Group 5 Insular Source: Own elaboration from Aena s tariff guides documents 8 Figure 2: Evolution of landing charge by airport group Group 1 Group 2 Group 3 Group 4 Group 5 Insular Source: Own elaboration from Aena s tariff guides documents

8 3. Econometric specification For the estimation of the multiproduct variable cost function we have chosen a flexible functional form, the multiproduct translog function with variables deviated from their geometric mean. It is based upon a second-order Taylor series expansion around the mean values: JJ MM JJ JJ ln VVVV iiii = αα + αα jj ln ww jjjjjj + ββ mm ln QQ mmmmmm γγ jjjj ln ww jjjjjj ln ww kkkkkk jj=1 mm=1 jj=1 kk=1 MM MM MM JJ δδ mmmm ln QQ mmmmmm ln QQ nnnnnn + λλ mmmm ln QQ mmmmmm ln ww jjjjjj mm=1 nn=1 mm=1 jj=1 JJ + θθ kk ln kk iitt θθ kkkk(ln kk iiii ) 2 + θθ jjjj ln kk iiii ln ww jjjjjj jj=1 MM + θθ mmmm ln kk iiii ln QQ mmmmmm + uu iiii mm=1 (1) where VVVV iiii is the total variable cost of airport ii in year tt, QQ mmmmmm is the amount of output mm for airport ii in year tt, ww jjjjjj is the variable input jj price for airport ii in year tt, and kk iiii is the quasifixed input of airport ii in year tt. We apply Shephard's Lemma to obtain the variable input cost shares to be estimated with the variable cost function in order to improve the efficiency of the estimation: JJ MM SS jjjjjj = ln VVVV iiii = αα ln ww jj + γγ jjjj ln ww kkkkkk + λλ mmmm ln QQ mmmmmm + θθ jjjj ln kk iiii + vv iiii jjjjjj kk=1 mm=1 (2) In order to fulfill the property of homogeneity of degree one for variable input prices in our cost specification, we impose the following restrictions: JJ JJ JJ JJ αα jj jj=1 = 1 ; γγ jjjj jj=1 = ; λλ mmmm jj=1 = ; θθ jjjj jj=1 = (3) Symmetry is imposed considering that: γγ jjjj = γγ kkkk, λλ mmmm = λλ jjjj, θθ jjjj = θθ kkkk. We might calculate airport marginal costs for each output by logarithmic differentiation of the variable cost function:

9 VVVV iiii MMMM mmmmmm = εε mmmmmm = ββ QQ mm + δδ mmmm ln QQ nnnnnn + λλ mmmm ln ww jjjjjj mmmmmm MM nn=1 JJ jj=1 + θθ mmmm ln kk iiii VVVV iiii QQ mmmmmm (4) As we have seen in the previous section, marginal costs are calculated in order to compare them with their corresponding airport charges set by regulators. The estimation of a variable cost function allows the calculation of the short-run scale economies SSrrrrrr, which are defined on the technology as the maximal proportionate growth rate of outputs, as all the variable inputs are expanded proportionally, given the level of installed capacity. SSSSSSSS iiii = VVVV(ww, QQ, kk) iiii QQ QQ mmmmmm mm mm = 1 mm εε mmmmmm (5) 4. Data We use a balanced panel data for 35 Spanish airports observed yearly from 29 until The economic and financial information was obtained from the annual reports published by AENA (various years), which provides guidance to individual Spanish airports in order to release accounting data that is both homogeneous and comparable across the airport system. Additional statistical data such as regional domestic product, population, have been gathered from Spanish National Statistical Institute (INE). The final panel data set consists of 21 observations. In a first stage of our analysis, we estimate a short run cost system of equations. To estimate this system, variables related to airports' variable cost, outputs, quasi-fixed inputs and input prices are required. Additionally, a set of control variables, and instruments have been also included to control for heterogeneity and endogeneity. Once the cost system has been estimated and marginal cost calculated, the second stage of the analysis compares Spanish airports' marginal costs with their corresponding passenger and landing charges. We describe next the variables used and the way in which they have been built. 5 In this study we have excluded the two heliports and eleven airports that are very small or do not operate regular flights (Albacete, Badajoz, Burgos, Córdoba, La Rioja, León, Madrid-Cuatro Vientos, Salamanca, Son Bonet and Huesca).

10 The variable cost (vc) is the sum of labor and intermediate consumption expenses. Regarding outputs, in the literature, it is common practice to consider the following ones: passenger, cargo, workload units (WLU), air transport movements (ATM) and the revenues from nonaeronautical or commercial activities. However, few studies used more than two or three of these outputs, Bottaso and Conti (212) is one of the few exceptions in airport cost analysis. This is because output variables present a high correlation among them. This study considers two outputs, passengers and the maximum take-off weight. On one hand, passenger output variable (pax) is measured as the number of loaded and unloaded passengers. On the other hand, construction of the maximum take-off weight variable is not straightforward. First, we obtain the number of landing operations by aircraft model in every airport and for every year from AENA (various years). Once we got these data, we need to know the mtow of each aircraft model by looking at the technical specifications of the aircraft. This is a very time consuming task and sometimes this information is hard to find. We have opted for a simpler procedure. We rank the aircraft models by number of operations and we obtain mtow data for the 3 or 4 most used aircraft models in each airport. For the rest of operations we use the average mtow of these most used aircrafts. Finally, we multiply every aircraft mtow data by the number of operations thus getting our output mtow variable. The inclusion of the maximum take-off weight as a proxy for landing services instead of ATM represents one of the novelties of our analysis. With respect to the other outputs considered in airport literature (cargo, workload units, and the revenues from non-aeronautical activity), their no-inclusion is due to the following reasons. First, cargo has been excluded of the analysis due to its contribution to Spanish airports' aeronautical revenues is small compared with other aeronautical services as passengers. However, there is an exception, Vitoria. This leads authors to include a dummy variable to control for this airport's particularity. Second, the number of workload units is usually considered as a proxy of aeronautical services, being highly correlated with the number of passengers. So the inclusion of passengers as an output makes not necessary to include this variable. Third, commercial revenues have not been included as an output because the airports' cost associated to these activities (opportunity cost of the area occupied, maintenance, consumptions) are not too much sensitive to an increase of their revenues (Santaló, 215). However, in order to control for differences in the level of commercial activities among airports, the share of commercial revenues in total airport revenues has been incorporated to the cost system. Other references in airport performance analysis literature that do not include non-

11 aeronautical revenues as an output are Gillen and Lall (1997), Sarkis (2), Pels et al. (21, 23), Abbott and Wu (22), Yoshida and Fujimoto (24), Martín-Cejas (25), Craig et. Al (25), Barros (28), Pathomsiri et al. (28), McCarthy (214), Kutlu and McCarthy (216), among others. On the input side, we have considered two variable and one quasi-fixed inputs.. The variable inputs are labor and intermediate consumptions, whereas the terminal surface (K) approximates the quasi-fixed one. We have computed Labor price (pl) as follows: (1) we obtain the annual cost per worker in the storage sector and support activities for transportation (in which airports are included) from the Annual Survey Labor Cost, published by the Spanish National Statistical Institute (INE). (2) We get a regional labor cost ratio by dividing the labor cost of each region by the national labor cost. (3) Finally, we get regional labor price by multiplying the annual cost per worker obtained in (1) by the ratio in (2). The regional-level Industrial Price Index is used as a proxy for airport intermediate consumption prices (pci). Additionally, we have included different control variables in the cost system in order to take into account heterogeneity among airports. First, a dummy variable for Vitoria airport (dv) which is highly specialized in cargo traffic. Second, the share of commercial revenues in total revenues in order to control airport heterogeneity regarding the level of commercial activity (shcom). Third, a group of dummy variables takes into account the different size of airports (g1-g5). Table 3: Descriptive statistics of the variables included in the cost system Variable Units Mean Std. Dev. Min Max Variable cost (vc) Mill. (21) Passengers (pax) Passengers E+7 Maximun take off weight Tons E+7 (mtow) Terminal surface (K) Square meters Labor price (pl) (21) Intermediate consumption prices (pci) Non-aeronautical revenue share (percom) Index Percentage Source: Own elaboration based on AENA's data.

12 5. Results 5.1. Short-run cost estimates We have estimated a short run cost system using three different specifications. In specification 1 we estimate the cost system using a 3SLS estimator, including time dummies to collect the effect of time and considering output variables endogenous. Specification 2 captures time effects using a trend variable. It includes non-neutral technological change, and exogenous outputs. Finally, the third specification includes non-neutral technological change as the previous one and, as in specification 1, outputs are taking as endogenous variables. Therefore, in specifications 1 and 3 we have included the GDP of the airports' region, landing and passenger charges, the density of motorways measured as km/1 squared-km in airport's region time, group and location dummies as instruments for output variables. In order to test the endogeneity of outputs, we carry out a Hausman over-identification test on models 1, and 3. In both cases, the null hypothesis of outputs being exogenous is rejected at the 5% level of significance. As we can see in Table 4, the specifications showed above fulfill the required regularity conditions: homogeneity of degree one on input prices, no decreasing in input prices and outputs and concavity in input prices 6. Additionally, in the three specifications the first order parameters exhibit the expected signs and are statistical significant with the exception of the quasi fixed input. Results also show the existence of scale economies. These scale economies are calculated as the inverse of the sum of the coefficient related to passengers and landing services. Additionally, the weight of intermediate consumptions in the variable cost is larger than labor. Specifications 2 and 3 show the existence of non-neutral technological change in inputs. On one hand, the parameters associated with the trend and its square are both negative and significant, indicating that over the years Spanish airport costs, ceteris paribus, have been reduced. Moreover, during the period studied, according to these specifications, the savings in variable costs related to the quasi-fixed input have increased over the time. The significance of the interaction between the trend and the quasi-fixed input also suggests the existence of some degree of short-run capital disequilibrium. 6 The procedure used for testing the concavity of the variable cost function in input prices follows the methodology of Baun and Linz (29).

13 Table 4: Short-run cost function estimates Specification 1 Specification 2 Specification 3 Coeffic. Std. Error Coeffic. Std. Error Coeffic. Std. Error cons *** *** *** pl *** *** *** pci *** *** *** K pax *** *** *** mtow *** ** ** plpl ** pcipci ** paxpax *** mtowmtow *** KK plpci *** plk pcik plpax *** ** *** plmtow *** *** *** pcipax *** ** *** pcimtow *** ***.95.2 *** Kpax Kmtow paxmtow ** g *** *** *** g *** *** *** g *** *** *** g *** *** *** percom dv *** *** *** t *** *** t *** *** tk ** ** tpl *** *** tpci.14.4 ***.14.4 *** tpax tmtow Y *** Y *** Y *** Y *** Y Notes: *, **, and *** denote parameters significant at 1%, 5% and 1%, respectively. Finally, it is possible to check that the importance of intermediate consumption costs in variable costs have increased; while in the case of labor, the opposite has occurred. This last result could imply a process of substitution between labor and intermediate consumption in the period

14 analyzed. This substitution process might be a consequence of the organizational change that AENA started in 212. Along this line, time effects in specification 1 show that from that year variable costs are lower than in the previous, ceteris paribus. In 212, AENA implement two strategic measures to make its airports more attractive to future investors. The first one is a reduction of its workforce of 11%. The second one is a plan to improve the efficiency of regional airports. These measures help AENA improve system's efficiency and reduce cost. Henceforth, the results analyzed are obtained from specification Marginal cost and economies of capacity utilization Table 5 displays the ECUs and the scale elasticities of passenger and landing services in the average of the groups of airports 2, 3, 4 and 5 described in section 2 and a additional group that includes insular airports,. On the other hand, annex 2 shows the same results in the average of each airport in the sample, individually. The passenger cost elasticity in the sample mean is.319. Table 5 shows that passengers cost elasticities increase with the level of passenger traffic. The highest passenger elasticities belong to hub airports, Madrid and Barcelona, which presents values of.628 and.62, respectively (see annex 2). On the other hand, the lowest values of passenger cost elasticities are presented in airports with levels of annual passenger traffic below 5. passengers as El Hierro (.223) and Valladolid (.235). On the other side, landing services cost elasticities are lower than those corresponding to passenger, many of them are not significant. This means that, in some airports, increasing in 1% the amount of tons of maximum take-off weight has not impact on variable costs. Thirdly, the mean of elasticities of capacity utilization (ECU) is 1.91, which implies scale economies in the short run. This result is higher than the value of 1.7 obtained by Martín et al. (211) for the Spanish airport system in the period This increase in the average ECU might be partially explained by the investments carried out in the first years of 2s. Additionally, it is possible to see in table 5 and annex 2 that all Spanish airports of the sample present scale economies in the short run and these scale economies decrease with the level of traffic which might imply a negative relationship between scale inefficiencies and airport size. This last result has been also shown by Bottaso and Conti (212) for the UK airport industry. Therefore, output cost elasticities and ECUs suggest is the existence of excess of capacity in those operations related to passenger and landing services. Then, it might be possible to increase the level of passenger and landing services keeping variable inputs constant. This means that Spanish airports' average cost might fall as more passenger and aircraft operation are served.

15 In spite of hub airports present the higher passenger cost elasticities, the highest marginal costs related to passengers correspond to regional airports. This result is explained by the fact that they present the highest passenger average costs. Table 5: Short-run cost elasticities for outputs and economies of capacity utilization Group Estimated cost elasticity Economies of passengers mtow Capacity Utilization *** *** 2.42 ***.227 ** *** 3.39 ***.19 *** *** ***.151 ** 2.1 *** *** *** Insular.333 ***.22 *** 1.81 *** Notes: *, **, and *** denote parameters significant at 1%, 5% and 1%, respectively. Table 6 reports values for average and marginal cost for each group. The estimated short run marginal costs of the sample are 9.39 and 2.51 euros per passenger and ton, respectively. Table 6 and annex 2 shows that, average costs are larger than marginal costs for all groups and airports so, as we have explained above, all Spanish airports would be operating in a region of increasing returns to scale. Regarding passengers, regional airports in group 5 present the highest levels of marginal costs, produced mainly by their high average costs. On the contrary, the main tourist airports and medium size airports aggregated in groups 2 and 3 present the lowest ones. With respect of the marginal costs associated a landing services, island group present the highest values. Specifically, Fuerteventura, Lanzarote, La Palma and Tenerife Sur have marginal costs over 5 euros per additional ton. Only the airports of Almería and Reus, belonging to group 4, present higher values that these airports. Group Table 6: Short-run average and marginal costs Estimated average cost Estimated marginal cost passengers mtow passengers mtow *** *** 2.82 ** *** *** *** ** *** Insular *** *** Notes: *, **, and *** denote parameters significant at 1%, 5% and 1%, respectively.

16 Figure 3: Evolution of passenger charges and marginal costs GROUP 1 GROUP 2 GROUP Pax MC Pax FEE Pax MC Pax FEE Pax MC Pax FEE GROUP 4 GROUP 5 INSULAR Pax MC Pax FEE Pax MC Pax FEE Pax MC Pax FEE Source: Own elaboration. Figures 3 and 4 present the evolution of established passenger and landing charges against the estimated short run marginal cost for both types of outputs (passengers and landing operations). Note that passenger charges are established above the marginal cost for all airport groups except for group 5, the one that account for the smallest airports. On the one hand, group 5 airports present the highest passenger marginal costs (7.81 on period average) and higher economies of capacity (2.664 on period average), this is, they are far below the optimum operational size. On the other hand, passenger charges set for these airports are notably lower compared to the rest of the groups. This fact may respond to a government strategy aimed at attracting low cost carriers to these airports.

17 Figure 4: Evolution of landing charges and marginal costs GROUP 1 GROUP 2 GROUP Mtow MC Mtow FEE Mtow MC Mtow FEE Mtow MC Mtow FEE GROUP 4 GROUP 5 INSULAR Mtow MC Mtow FEE Mtow MC Mtow FEE Mtow MC Mtow FEE Source: Own elaboration. 6. Conclusions and policy implications This article analyses marginal costs and aeronautical airport charges for the Spanish airports in a pre-privatization period (29-214). We propose a flexible short-run cost function system to estimate marginal costs for passengers and aircraft operations. Our results show the existence of non-neutral technological change for the period from 29 until 214. A possible explanation might be the institutional and regulatory changes with regard to the different privatization proposals for the Spanish airport system. We also suggest the existence of some degree of short-run capital disequilibrium motivated by the overcapitalization of Spanish airports. We demonstrate that the evolution of airport charges does not follow the evolution of marginal costs. Moreover, aeronautical airport charges are set above the short-run marginal costs with the exception for the smallest airports, aggregated in group 5. The comparison of airport charges and airport marginal costs for a more recent period in which the privatization process had already achieved, the inclusion of non-aeronautical activities following the literature of two-sided markets or the estimation of airport charge elasticities of both aeronautical and non-aeronautical demands might be some issues for further research

18 7. References Abbott, M., & Wu, S. (22). Total factor productivity and efficiency of Australian airports. Australian Economic Review, 35(3), Barros, C. P. (28). Technical efficiency of UK airports. Journal of Air Transport Management, 14(4), Baun, C.F. & Linz, T. (29) Evaluating concavity for production and cost functions. The Stata Journal, 9(1), Bel, G. & Fageda, X. (211a) Privatization, regulation and airport pricing: an empirical analysis for Europe. Journal of Regulatory Economics, 37, Bel, G. & Fageda, X. (211b) La reforma del modelo de gestión de aeropuertos en España: Gestión conjunta o individual?. Hacienda Pública Española/Review of Public Economics, 196(1), Bottasso, A. & Conti, M. (212) The cost structure of the UK airport industry. Journal of Transport Economics and Policy, 46(3), Bottasso, A., & Conti, M. (217). The Cost Structure of the Airport Industry: Methodological Issues and Empirical Evidence. In The Economics of Airport Operations, Emerald Publishing Limited. Bilotkach, V, Clougherty, J.A, Mueller, J. & Zhang, A. (212) Regulation, privatization, and airport charges: panel data evidence from European airports. Journal of Regulatory Economics. CNMC (214) El sector aeroportuario en España: Situación actual y recomendaciones de liberalización. E/CNMC/2/14 Report. Czerny, A.I. (26) Price-cap regulation of airports: single-till versus dual-till. Journal of Regulatory Economics, 3, Hidalgo-Gallego, S., Martínez-San Román, V., & Núñez-Sánchez, R. (217). Estimation of Allocative Efficiency in Airports for a Pre-Privatization Period. In The Economics of Airport Operations, Emerald Publishing Limited. Ley 21/23, de 7 de julio, de Seguridad Aérea. BOE núm. 162, de 8/7/23. Ley 18/214, de 15 de octubre, de aprobación de medidas urgentes para el crecimiento, la competitividad y la eficiencia. BOE núm. 252, de 17 de octubre de 214

19 Lu, C. & Pagliari, R. (24) Evaluating the potential impact of alternative airport pricing approaches on social welfare. Transportation Research Part-E: Logistics and Transportation Review, 4(2), Kutlu, L., & McCarthy, P. (216). US airport ownership, efficiency, and heterogeneity. Transportation Research Part E: Logistics and Transportation Review, 89, McCarthy, P. (214). US airport costs and production technology a translog cost function analysis. Journal of Transport Economics and Policy (JTEP), 48(3), Martín-Cejas, R. R. (25). Two-step estimation method for translog cost function: An application to Spanish airport networks. International Journal of Transport Economics/Rivista internazionale di economia dei trasporti, Martín, J.C., Román, C. & Voltes-Dorta, A. (211) Scale economies and marginal costs in Spanish airport. Transportation Research Part E: Logistics and Transportation Review, 47, Mathisen, T. A., Jørgensen, F., & Solvoll, G. (214). Marginal costs pricing of airport operations in Norway. Research in Transportation Economics, 45, Núñez-Sánchez, R. (213). Marginal costs, price elasticities of demand, and second-best pricing in a multiproduct industry: An application for Spanish port infrastructure. Journal of Transport Economics and Policy, 47(3), Núñez-Sánchez, R. (215). Regional public support to airlines and airports: An unsolved puzzle. Transportation Research Part E: Logistics and Transportation Review, 76, Pathomsiri, S., Haghani, A., Dresner, M., & Windle, R. J. (28). Impact of undesirable outputs on the productivity of US airports. Transportation Research Part E: Logistics and Transportation Review, 44(2), Pels, E., Nijkamp, P., & Rietveld, P. (21). Relative efficiency of European airports. Transport Policy, 8(3), Pels, E., Nijkamp, P., & Rietveld, P. (23). Inefficiencies and scale economies of European airport operations. Transportation Research Part E: Logistics and Transportation Review, 39(5), Real Decreto-ley 13/21, de 3 de diciembre, de actuaciones en el ámbito fiscal, laboral y liberalizadoras para fomentar la inversión y la creación de empleo.

20 Real Decreto-ley 2/212, de 13 de julio, de medidas para garantizar la estabilidad presupuestaria y de fomento de la competitividad. Santaló, J. S. (215). El impacto de la nueva regulación aeroportuaria sobre las tasas aéreas y el conflicto entre la Comisión Nacional de Mercados y Competencia (CNMC) y AENA a cuenta de la misma. Papeles de economía española, (145), Sarkis, J. (2). An analysis of the operational efficiency of major airports in the United States. Journal of Operations management, 18(3), Yoshida, Y., & Fujimoto, H. (24). Japanese-airport benchmarking with the DEA and endogenous-weight TFP methods: testing the criticism of overinvestment in Japanese regional airports. Transportation Research Part E: Logistics and Transportation Review, 4(6),

21 Annex 1. Definition of variables related to airport charges In order to compare landing and passenger marginal costs with their respective charges, these have been estimated following the next procedures. Passenger charge AENA (29) establishes passenger charge (ppax) as "providing passenger departure areas not accessible to visitors airport terminal areas, as well as complementary airport facilities." Payment must be performed by airlines or government. However, passengers usually pay through this charge as an increase in the ticket fare. The amount for this charge is determined by the origin and destination of the flight between 29 and June 211: a) between islands within the same archipelago (e.g. Tenerife Sur-Las Palmas de Gran Canaria); b) from any airport located in an island to any Spanish airport (e.g. Ibiza-Santander); c) from any peninsular airport to any airport located in the European Economic Area (EEA); d) from any Spanish airport to any airport non-located in the EEA. A new airport charge scheme, implemented since June 211, allows different airport charges according the volume of passengers of each airport. In this sense, the new regulatory system increase passenger charges for Madrid, Barcelona and airports specialized in tourism whereas decreases them for regional airports. This scheme discriminates between destinations in the EEA and international passengers. As in the previous regulation, it considers routes in the same archipelago or routes from any airport located in an island to any another Spanish airport. In this study, we proceeded to the calculation as follows: (1) we have identified the type of passengers for each airport. (2) It has made a weighted average percentage of passengers of each type by the corresponding charge, obtaining an aggregated passenger charge for each airport and year, measured in euros per passenger. Landing charge The average landing charge per ton (pmtow) has been computed as a weighted average of the different landing charges set by AENA, based on the airport group, the type and class of flight and the acoustic classification of the aircraft. Given that there is no information available at source-destination-type of aircraft level, the average landing charge per airport per year will be an approximation, although, as far as we know, it seems to be the best approximation in the literature yet. In 29 and 21 there were different landing charges at the Spanish airports depending on whether the flight came from the European Economic Area (EEA), as well as on

22 the weight of the aircraft. From 211, this differentiation disappeared and the landing charge does not depend on the origin country of the flight or the weight of the aircraft. Only for the case of island airports AENA's establishes different charges depending on whether on whether the flight originates in another island of the same region or in islands of other Spanish regions. Given the above-mentioned lack of information availability, we have decided to make the following decisions in order to obtain an average landing charge for each airport. First, we do not take into account the acoustic classification of the aircraft due to the lack of data or the extreme complexity to achieve them. Second, for 29 and 21, where there are differentiated charges according to the weight of the aircraft, we have opted to compute the corresponding charge for aircrafts with a weight between 1 and 1 tons (the majority of aircrafts in Spanish airports). Finally, for all the airports in 29 and 21 and for the island airports throughout the period we have calculated the percentage of landings that originate from EEA countries, from outside the EEA, from islands located in other Spanish region and from islands located in the same Spanish region. This allows us to calculate the weights we need to compute the average landing charge for each of Spanish airports. Table A1 shows the descriptive statistics of the variables related to airport charges. Table A1: Descriptive statistics of the variables related to airport charges Variable Units Mean Std. Dev. Min Max Landing charge Euros (21) Passenger charge Euros (21) Source: Own elaboration based on AENA's data.

23 Annex 2. Spanish airports marginal costs and economies of capacity utilization Table A2: Cost elasticity, marginal cost and ECUs by airport Estimated cost elasticity Economies of capacity Estimated marginal cost passengers mtow utilization passengers mtow Alicante.457 *** *** *** Almería.249 ***.237 ** 2.56 *** *** 7.68 ** Asturias.47 *** *** 4.6 ***.428 Barcelona.62 *** *** *** 1.8 Bilbao.522 *** *** ***.427 Coruña.42 *** *** 4.91 *** Fuerteventura.257 **.335 ** *** ** ** Gerona.429 *** *** *** Granada.341 ***.151 ** 2.36 *** *** 3.32 ** Hierro.223 ** *** ** 8.1 Ibiza.413 ***.189 ** *** *** ** Jerez.332 ***.165 *** 2.13 *** 4.21 *** *** Lanzarote.311 ***.29 *** *** *** *** La Palma.291 ***.213 *** *** *** *** La Gomera.355 *** *** *** Gran Canaria.37 ***.27 *** *** 1.98 *** *** Madrid.628 *** *** 5.14 *** Menorca.258 **.294 *** *** 2.73 ** *** Málaga.493 *** *** *** 1.99 Melilla.315 *** *** 6.3 *** 3.75 Mallorca.485 *** *** *** Pamplona.356 *** *** 1.4 *** Reus.242 **.264 ** *** ** ** San Javier.33 ***.186 *** *** 2.15 *** *** San Sebastian.353 *** *** *** Tenerife Sur.296 **.326 *** 1.66 *** ** *** Tenerife Norte.399 ***.188 ** 1.74 *** *** ** Santander.366 ***.135 ** *** 2.87 *** ** Santiago.382 ***.169 ** *** *** ** Sevilla.453 *** *** *** Valencia.473 *** *** 3. *** Valladolid.235 **.21 * *** ** 8.41 * Vigo.44 *** *** ***.92 Vitoria.391 *** *** 1.69 *** Zaragoza.355 ***.11 * *** *** * Notes: *, **, and *** denote parameters significant at 1%, 5% and 1%, respectively.